13 posts tagged with "llm"

In our ongoing quest to improve market comparison reliability, we've discovered a powerful new approach: multi-turn conversations with LLMs. This technique breaks down complex market analyses into a series of focused questions, leading to more accurate and explainable trading decisions.

The Journey: How This Strategy Evolved​

Origins and Initial Exploration​

This project began with an analysis of the original blog post on crypto news sentiment analysis, which outlined three fundamental approaches to quantifying news impact on cryptocurrency markets:

1. Source-Specific Impact Measurement: Tracking specific influencers/sources and measuring their historical market impact
2. Topic-Based Sensitivity Analysis: Categorizing news by topic and measuring market sensitivity to different categories
3. Narrative Lifecycle Analysis: Treating market narratives as living entities with lifecycles, tracking how they emerge, peak, and fade

The post expanded these into three implementation methodologies:

1. Multi-Source Sentiment Analysis with Historical Correlation
2. Topic-Based Market Impact Analysis
3. Network Effect & Propagation Analysis

In our quest to improve market comparison accuracy, we've experimented with different versions of sorting prompts. This analysis compares their effectiveness and helps identify the optimal approach.

Test Setup​

• Markets: BTCUSDT vs ETHUSDT
• Timeframe: 4h
• Core Indicators: RSI, MACD, Bollinger Bands, Volume, ADX, Volatility, Ichimoku
• Models Tested: LLAMA3.2:3B, GEMMA2:2B, DEEPSEEK:R1:1.5B